FEDERAL STATISTICS:

Report to Congressional Requesters

June 2026

GAO-26-107538

United States Government Accountability Office

A report to congressional requesters

Contact: Thomas Costa at costat@gao.gov or Michael Hoffman at hoffmanme@gao.gov.

What GAO Found

The Bureau of Labor Statistics’s (BLS) Employment Situation report (the Jobs Report) provides key information on the nation’s economy based on data from two surveys—one of households (the household survey) and one of employers (the establishment survey). Stakeholders with jobs data expertise said the report generally meets users’ needs. However, they said occasional large revisions can make the data less useful for informing timely decisions, and BLS faces risks to data quality due to lower survey response rates over time (see figure). BLS met its goals for the data’s precision and size of the revisions from fiscal years 2020 through 2025, but it relaxed one goal during the COVID-19 pandemic.

BLS obtains input from data users and technical experts through various means to inform improvements to the Jobs Report. However, it does not have a plan to address gaps in obtaining regular user and technical feedback since the Departments of Commerce and Labor eliminated three advisory committees in 2025. Having a plan would help ensure that changes to the Jobs Report meet users’ needs and address technical challenges while preserving data quality.

BLS and the Census Bureau—which conducts the household survey for BLS—have met federal requirements for assessing the effects of lower survey response rates on the Jobs Report’s accuracy but released limited information on their findings. Studies by agency staff found limited effects of lower response rates on the accuracy of the household survey data. However, BLS has not released its assessments of the establishment survey due to limitations, including challenges linking the survey data with more comprehensive sources. BLS has started a new establishment survey study designed to address earlier limitations. Releasing the results could help users maintain trust in the data’s accuracy.

Agency efforts and additional options identified by stakeholders aim to enhance the quality of the Jobs Report data. For example, BLS and Census are planning to add an online response method to the household survey in 2027 to try to increase response rates and lower data collection costs. However, BLS and Census officials said recent funding constraints may delay full implementation. Also, stakeholders, researchers, and BLS officials identified additional options to address survey challenges. For instance, some suggested incorporating administrative data from private and other government sources (such as unemployment insurance data) to help mitigate lower survey response rates.

Why GAO Did This Study

Economic policymakers, businesses, and others rely on the monthly Jobs Report for accurate and timely information on the health of the nation’s economy, such as the unemployment rate. However, in recent years, questions have been raised about the quality of the data in the report.

GAO was asked to review the quality of the data in the Jobs Report. GAO’s report addresses (1) the extent to which the Jobs Report data meet users’ needs for accurate, useful, and timely information, and challenges BLS faces in producing these data, (2) the extent to which BLS and Census have followed federal requirements for assessing survey response rates and communicating findings to data users, and (3) potential options to address survey challenges and associated considerations.

GAO reviewed data on BLS’s performance goals for the quality of the Jobs Report data. GAO also reviewed relevant government and academic studies and agency documentation. GAO interviewed agency officials and 14 stakeholders that were selected to provide a range of expert perspectives on the quality of the data for informing economic policy and business decisions and on options to address survey challenges. GAO compared BLS and Census processes to federal statistical policies and relevant federal standards for internal control.

What GAO Recommends

GAO is making three recommendations, including that BLS develop a plan to address gaps in its ability to obtain external input on the Jobs Report data and publish an assessment on the effects of survey nonresponse on the establishment survey. BLS generally agreed with the recommendations.

 

 

 

 

BLS                                         Bureau of Labor Statistics

COVID-19                               coronavirus disease 2019

establishment survey              Current Employment Statistics program

household survey                   Current Population Survey

Jobs Report                            Employment Situation report

OMB                                       Office of Management and Budget

June 2, 2026

The Honorable Tim Walberg
Chairman
Committee on Education and Workforce
House of Representatives

The Honorable Virginia Foxx
House of Representatives

The Bureau of Labor Statistics’s (BLS) Employment Situation report—commonly referred to as the Jobs Report—is a widely used source for data on national employment trends. It provides critical information on the health of the nation’s economy. The report includes data on key economic indicators from two surveys. The Current Population Survey (the “household survey”) is conducted by the Census Bureau on behalf of BLS and provides the national unemployment rate. The Current Employment Statistics program (the “establishment survey”) is conducted by BLS and provides data on payroll employment growth—the number of jobs added to the economy or lost.

Economic policymakers, business leaders, and others rely on the agency’s monthly Jobs Report for accurate and timely information that they can use to inform decisions. For example, Federal Reserve policymakers closely monitor data on the national unemployment rate and employment growth when determining interest rate levels to manage inflation and promote maximum employment in the economy. Businesses use the data to inform workforce and investment decisions, such as whether to hire more workers or expand operations.[1]

In recent years, questions have been raised about the quality of the data in the Jobs Report. For example, there have been concerns that the Jobs Report estimates may miscount the number of jobs added or lost when there are significant shifts in the direction of the economy, such as at the beginning of a recession or expansion.

You asked us to review the quality of the data in the Jobs Report. This report examines (1) the extent to which the employment data in the Jobs Report meet users’ needs for accurate, useful, and timely information and challenges BLS faces in producing these data; (2) the extent to which BLS and Census have followed federal requirements for assessing survey response rates and for communicating the findings of these assessments to data users; and (3) what potential options BLS, Census, stakeholders, and research have identified to address survey challenges, and what considerations are associated with each option.

We assessed the accuracy, usefulness, and timeliness of the Jobs Report data based on relevant dimensions from the Federal Committee on Statistical Methodology data quality framework. This framework identifies common definitions, objectives, and best practices for federal agencies to follow for producing and reporting data.[2] Specifically, we assessed whether BLS met its performance goals for accuracy, usefulness, and timeliness for the household and establishment survey data from fiscal years 2020 through 2025.[3] We also analyzed survey response rates and information that BLS uses to evaluate its performance goals, including data revisions and measures of survey precision. We assessed the reliability of these data by reviewing relevant documentation, interviewing knowledgeable officials, and conducting electronic data testing. We found the performance data to be sufficiently reliable for determining the extent to which BLS met its own accuracy, usefulness, and timeliness goals and reviewing recent survey response rate trends.[4] Moreover, we assessed the extent to which BLS’s processes for obtaining user and expert input on the Jobs Report data met BLS and Office of Management and Budget (OMB) policies and federal standards for internal control.[5]

We also interviewed 14 stakeholders with expertise on the data and methods in the Jobs Report to obtain their views on the accuracy, usefulness, and timeliness of the employment data for informing economic policy and business decisions.[6] These interviews included discussion of the specific strengths and limitations of the household and establishment surveys, including current and emerging data quality challenges (which we refer to as “survey challenges”) faced by data users, BLS, and Census. Specifically, we selected a nongeneralizable sample of 14 stakeholders from the following categories: (1) public policy decision-makers and advisors, (2) banks and investment firms, (3) business associations, (4) former BLS commissioners, and (5) researchers with expertise related to the household or establishment survey. We selected these stakeholders to obtain a broad range of institutional perspectives, but their views may not be representative of the stakeholder population as a whole.

To further understand how survey challenges may have affected the accuracy of the Jobs Report data, we conducted a literature review. Specifically, we reviewed government and academic studies that examined the accuracy of the employment data in either the household survey or the establishment survey. We included studies that were published from January 2020 to September 2025.[7]

To assess the extent to which BLS and Census have followed federal requirements for assessing survey response rates and for communicating the findings of these assessments to data users, we reviewed BLS and Census data and documentation. Specifically, we reviewed survey response rates that BLS and Census publish, as well as internal and public BLS and Census assessments of the effects of response rates on data quality.[8] We also interviewed BLS and Census officials to understand their processes for reviewing, assessing, and communicating information on response rates to data users. We evaluated the extent to which these processes met relevant OMB and BLS policies and relevant internal control standards.[9]

We interviewed BLS and Census officials, as well as our selected stakeholders, to identify agency initiatives and potential options to address survey challenges. We also identified potential options to address survey challenges through our literature review. In our interviews with agency officials and stakeholders, we gathered information on the anticipated effectiveness of the potential options for addressing survey challenges. In addition, we asked about the feasibility, costs, effects on data quality, and other considerations associated with each option, as appropriate. We also reviewed relevant agency research and documentation, as well as research shared by stakeholders, to further understand these options and key considerations.

To corroborate information shared by agency officials and stakeholders, we also analyzed BLS budget data for the household and establishment surveys from fiscal years 2020 through 2025. We assessed the reliability of these data by interviewing knowledgeable officials. We found the budget data to be sufficiently reliable for describing trends over this period. For more information on our methodology, see appendix I.

We conducted this performance audit from April 2024 to June 2026 in accordance with generally accepted government auditing standards. Those standards require that we plan and perform the audit to obtain sufficient, appropriate evidence to provide a reasonable basis for our findings and conclusions based on our audit objectives. We believe that the evidence obtained provides a reasonable basis for our findings and conclusions based on our audit objectives.

Background

Overview of the Jobs Report

BLS prepares the Jobs Report using data from two statistical surveys (see fig. 1). The household survey provides the national unemployment rate—an estimate of the percentage of the U.S. labor force searching for work—and other labor force indicators.[10] Through in-person or telephone interviews, Census collects these data from a nationally representative statistical sample of households on behalf of BLS. The establishment survey provides total payroll employment growth, as well as data on workers’ hours and their earnings.[11] BLS collects these data from a nationally representative statistical sample of employers through phone interviews and via an online web-reporting system.[12] Data from households and employers are collected monthly. BLS analyzes the data from both sources and then issues the Jobs Report on the first Friday of each month.[13]

^aHousehold survey refers to the Current Population Survey.

^bEstablishment survey refers to the Current Employment Statistics program.
Note: Establishment survey data do not include employees in agriculture, in private households, or the self-employed.

Each report presents labor market data for the previous month and recent trends. The report presents these data for the nation as a whole and for various demographic groups and industries. BLS also publishes more detailed subgroup data that users can download from its website after the report is released.

Each report also includes estimates of certain labor market data with adjustments for regular seasonal patterns (“seasonally adjusted data”). Seasonal adjustments are designed to remove the effects of regularly occurring fluctuations in the labor force, such as those due to changes in the weather and major holidays. These fluctuations can be large, such as the number of youth who temporarily enter the labor force each summer, according to BLS documentation. Seasonally adjusted data that remove such patterns can make it easier to distinguish other trends in economic activity.

A broad range of data users rely on the information published in the Jobs Report to assess employment conditions and the broader economy (see fig. 2).

Revisions to the Jobs Report Data

In keeping with longstanding practice, BLS updates the establishment survey data in the Jobs Report over time through a number of “revisions” to allow employers additional time to respond to the survey and to incorporate more complete workforce data.[14] There are two types of revisions—monthly revisions and annual benchmark revisions. BLS has used this combination of monthly revisions and annual benchmarking to update the data since 1982.

Monthly Revisions

Each month, BLS releases initial estimates of payroll employment, hours, and earnings from the establishment survey and then subsequently revises these data two more times. This is because some employers do not report their employment information to BLS in time for the initial release of the Jobs Report data. The establishment survey collects data for payrolls that include the 12th day of the month.[15] BLS starts collecting data for the survey on the first business day after the 12th and continues to do so until the Monday before the Jobs Report’s release. Therefore, the initial estimates in the Jobs Report are based on the sample of employers who reported their payroll information within that window.[16] BLS then continues to collect information from the employers who did not report their payroll information in time and includes this information for the month’s second and third releases, revising its estimates for each release based on this additional information. For example, initial estimates for January are published in February. A revised estimate for January’s data is published in March. The final monthly revision of January’s data is published in April.

These revisions allow BLS to incorporate data from a larger proportion of employers over time. For example, in 2024 BLS collected additional responses that increased the survey’s coverage of employers by about 30 percentage points on average—with responses rising on average from about 60 percent in the first release to about 90 percent by the third release.[17]

Annual Benchmark Revisions

Each February, BLS revises the establishment survey data based on a comparison to comprehensive employment data in a process called the “annual benchmark revision.” Like the monthly revisions, the goal of the annual benchmark revision is to provide more accurate data on employment (i.e., the number of jobs). The establishment survey’s original estimates for the Jobs Report are based on a statistical sample of employers that cover about 26 percent of total U.S. employment, according to BLS documentation. BLS’s annual benchmark revision is based on comprehensive administrative employment records that it obtains quarterly from state agencies through their unemployment insurance programs. These data cover about 97 percent of total employment. BLS uses these administrative employment records to calculate a definitive count of employment for the previous March—the benchmark.[18]

BLS releases its annual benchmark revisions in February. BLS revises 21 months of previously published estimates with the benchmark release. Specifically, BLS revises estimates for the 11 months prior to the benchmark, and it revises the 9 months that follow the benchmark.[19] BLS also typically publishes a preliminary estimate of the annual benchmark revision in August, although it does not revise its previously published estimates with this preliminary estimate.[20] BLS considers the degree of change associated with the annual benchmark revision to the establishment survey to be a measure of overall error in the establishment survey. This is because the revision is based on a comparison between the original establishment survey estimates and employment data that cover nearly the entire U.S. workforce.

Figure 3 shows a recent example of BLS’s monthly and annual benchmark revisions, based on employment data originally released in the Jobs Report for July 2023.[21] As a result of two monthly revisions and two annual benchmark revisions, BLS revises the establishment survey estimates at least four times.[22] For example, the initial survey response for July 2023 showed U.S. payroll employment grew by 187,000 jobs. BLS reported this initial estimate in August 2023. This growth was revised in the following two months, first downward to an increase of 157,000 jobs in September 2023 and then upward to an increase of 236,000 jobs in October 2023. Once administrative employment data were available in February 2024 for the March 2023 benchmark, July 2023 job growth was revised to 184,000 jobs. With additional administrative data in February 2025 for the March 2024 benchmark, July 2023 job growth was revised to 148,000 jobs. The range of these revisions is discussed in more detail below.

Note: Estimates of employment growth and revisions have been seasonally adjusted. Prior to the annual benchmark revision, estimates of employment growth are based on surveys and have a margin of error of about 130,000 at the 90 percent confidence level. We selected July 2023 to illustrate BLS’s revision process because its monthly revision for July 2023 was close in magnitude to the overall average for that year and the revision was during the latest year in which 2 full years of annual benchmark revisions were available at the time of our analysis.

^aEstablishment survey refers to the Current Employment Statistics program.

Federal Requirements and Guidance Related to Data Quality

As principal statistical agencies, BLS and Census are expected to follow federal laws, standards, and guidelines to ensure that statistical policy is carried out consistently and effectively across the federal government. For example, Office of Management and Budget (OMB) policies require statistical agencies to produce data that are accurate, useful, and timely and set standards for running statistical surveys.[23]

When assessing whether the Jobs Report and other data products are accurate, useful, and timely, BLS policy applies the Federal Committee on Statistical Methodology data quality framework.[24] According to the framework, the accuracy of survey data—such as estimates of national unemployment or employment growth—can be affected by multiple factors. The accuracy of such estimates depends on how close they are to their true value (e.g., the value that BLS would compute based on information from the entire population of U.S. households or employers instead of only a statistical sample). For example, estimates that are computed from larger samples will generally have less statistical error than estimates computed from smaller samples, and are therefore more likely to be accurate. In addition, important differences between survey samples and the larger populations they are meant to represent can push—or bias—an estimate consistently in one direction away from the true value depending on the nature of the differences between survey participants and nonparticipants.

Beyond accuracy, the framework also identifies five additional dimensions of data quality that together determine the usefulness of federal data products like the Jobs Report, including their timeliness:[25]

·         Relevance. Whether the product is targeted to meet data users’ needs.

·         Granularity. The amount of detail available for key data elements.

·         Accessibility. The ease with which data users can obtain products and documentation.

·         Timeliness. The length of time between the period that the product describes and when it is available to data users.

·         Punctuality. Whether the product is released on schedule.

Annual Agency Performance Goals Related to Data Quality

BLS sets goals annually for monitoring the accuracy, usefulness, and timeliness of the Jobs Report data and other data products as part of the budget development and justification process. For example, to monitor the timeliness of the Jobs Report data, BLS tracks whether it releases its estimates from each survey in the Jobs Report on time each month according to an OMB approved schedule. As part of this process, these goals are reviewed and assessed each year by Department of Labor and OMB officials, according to BLS officials. BLS sets separate goals for the Jobs Report data for the household and establishment surveys.

Stakeholders with Jobs Data Expertise Said the Report Generally Meets Users’ Needs, but BLS Does Not Have a Plan to Address Gaps in External Input

Stakeholders we interviewed generally agreed that, overall, BLS’s Jobs Report meets users’ needs for accurate, useful, and timely data for informing economic policy and business decisions but also noted risks to maintaining data quality. From fiscal years 2020 through 2025, BLS met its own accuracy goals for key estimates in the Jobs Report, such as the size of BLS’s revisions to the payroll employment data from the establishment survey, although it temporarily lowered one goal during the COVID-19 pandemic. Nevertheless, lower survey response rates, occasional large revisions, and other survey challenges can, at times, affect the accuracy and usefulness of the data. BLS has obtained some input from data users and technical experts regarding the Jobs Report data, which it takes into account when considering changes, such as initiatives to address lower survey response rates. However, BLS does not have a plan to address gaps in obtaining regular external feedback from data users and technical experts to improve Jobs Report data. Gaps exist since the Departments of Labor and Commerce eliminated three advisory committees that BLS previously relied on for feedback.

Stakeholders Said Data Users Can Rely on Jobs Report Estimates to Inform Policy and Business Decisions

How Did GAO Select Stakeholders to Interview? We selected stakeholders with expertise in the Employment Situation report (Jobs Report) data and methods to obtain a range of institutional perspectives. We interviewed a nongeneralizable sample of stakeholders from the following categories: ·       Public policy decision-makers and advisors ·       Banks and investment firms ·       Business associations ·       Former Bureau of Labor Statistics commissioners ·       Researchers with methodological expertise Their views may not be representative of the broader stakeholder population. For more information on GAO’s stakeholder selection methodology, see appendix I. Source: GAO.  |  GAO‑26‑107538

The stakeholders we interviewed generally agreed that the Jobs Report data are sufficiently accurate, useful, and timely for informing economic policy and business decisions. For example, all 14 of the stakeholders we interviewed agreed that BLS’s analysis and methods for the report are based on accepted statistical practices, that the data in the report meet users’ needs, and that the information is sufficiently timely for informing economic policy and business decisions (see fig. 4).

Note: Stakeholders rated their level of agreement or disagreement to questions related to accuracy, usefulness, and timeliness of the Jobs Report data for informing economic policy and business decisions. We used minor variations in the wording of some statements based on stakeholder type and present the results using standardized wording in this figure. For additional details on this methodology and the stakeholder responses, see appendix I.

Twelve of the 14 stakeholders agreed the report’s overall accuracy and reliability are acceptable for informing economic policy and business decisions. Many stakeholders cited aspects of the design of the household and establishment survey as particular strengths of the data.[26] For example, they noted the large number of households and employers that Census and BLS survey each month and that each survey is nationally representative.[27] Citing these features, some stakeholders said the household and establishment surveys are considered “the gold standard” for surveys. However, two stakeholders neither agreed nor disagreed, citing the potential adverse effects of lower survey response rates, the need for more precise survey estimates, and the need for smaller data revisions. We discuss these topics in more detail below.

Many stakeholders we interviewed said they use the Jobs Report data to gain insights into employment trends. For example, some stakeholders said the ability to use the household survey to analyze employment trends over time, including over the business cycle, is a strength. There have been relatively few changes to the questions in the household survey over time, according to BLS documentation, and users can download historical data on key employment measures such as the monthly national unemployment rate back to 1948.

Thirteen of 14 stakeholders agreed that the information in the Jobs Report and BLS’s accompanying technical documentation is adequate for understanding key features and uses of the data. They also agreed that the data and documentation can be obtained in understandable forms and formats. However, one stakeholder disagreed with these statements, explaining that some data users have difficulty finding information they need on the BLS website.

All 14 stakeholders agreed that the Jobs Report is sufficiently timely to meet users’ needs. Both surveys in the Jobs Report followed OMB’s timeliness requirement for federal economic indicators, which states that agencies should compile and release survey estimates no later than 22 working days after the end of the period they cover.[28] BLS schedules the release of the Jobs Report approximately 3 weeks after the period each survey covers.[29] Moreover, BLS released the report on schedule each month from fiscal years 2020 through 2025.[30] One stakeholder mentioned that the Jobs Report is one of the earliest economic indicators that their institution receives on the state of the national economy for the previous month when evaluating policy options.

However, some stakeholders said that the monthly frequency can be inadequate for monitoring sudden changes in labor market conditions. For example, one stakeholder said that business closures in March 2020 immediately following the beginning of the COVID-19 pandemic were not fully captured until the May release of the Jobs Report, delaying key information needed for policymakers’ response to the pandemic.

Jobs Report Estimates Have Met Agency Accuracy Goals

We found that BLS met its annual accuracy goals for the survey estimates in the Jobs Report from fiscal years 2020 through 2025 amid data collection challenges.

Household survey accuracy goal. For the household survey, BLS’s accuracy goal is based on the survey’s precision requirements for its unemployment rate estimates. Specifically, the goal is that a monthly change in the national unemployment rate of at least 0.2 percentage points should be statistically significant—an indication that the change is likely not due to chance. BLS measures this goal using the margin of error.[31] From fiscal years 2020 through 2025, BLS met this goal for the household survey every month.

BLS officials said that it will be difficult to meet this goal in the future. Specifically, BLS expects that lower response rates to the household survey will make it challenging to maintain this level of precision. Consequently, the agency relaxed its precision requirement for the national unemployment rate in April 2025.[32] The agency also relaxed the corresponding accuracy goal for fiscal year 2026. We discuss this change and its implications for users in more detail below.

Establishment survey accuracy goals. For the establishment survey, BLS has accuracy goals for the size of its revisions to payroll employment. The revision process is designed to provide users with initial estimates soon after the month is over, followed by revised estimates as more complete workforce data become available. These revision goals help BLS monitor the survey’s accuracy based on how much the estimates change as a result of obtaining more complete data. BLS has separate goals for its (1) monthly revisions and (2) annual benchmark revisions.

We found that, overall, BLS met its accuracy goals for both types of revisions from fiscal years 2020 through 2025. However, BLS temporarily lowered its goal for monthly revisions during the COVID-19 pandemic, and some revisions exceeded typical levels.

·         Accuracy goal for monthly revisions to estimates of payroll employment growth. BLS’s goal is that the monthly revisions must be within 0.1 percentage points of its initial estimate in 10 out of 12 months in a given year. BLS measures this goal using the size of the revision to employment growth from the initial release to the third release.[33] From fiscal years 2022 through 2025, BLS met this accuracy goal (see fig. 5). BLS also met its goal for the monthly revisions in fiscal years 2020 and 2021, but it temporarily lowered its goal during this period from 10 to 8 months of the year. It did so to accommodate larger revisions during the COVID-19 pandemic due to data collection issues and economic fluctuations.[34]

Note: Revisions are based on the comparison of initial and third releases of nonseasonally adjusted payroll employment growth in the Jobs Report. BLS monitors monthly revisions published during the federal fiscal year. These correspond to revisions to employment data from July to June each year. BLS generally has a performance goal of monthly revisions within a threshold of 0.1 percentage points for at least 10 months of each fiscal year. The COVID-19 period shown in the graphic does not reflect the start and end dates of the public health emergency but rather the time period from January 2020 through June 2021 in which BLS lowered this performance goal due to larger expected revisions.

^aEstablishment survey refers to the Current Employment Statistics program.

·         Accuracy goal for annual benchmark revisions to payroll employment estimates. BLS’s goal is that annual benchmark revisions to these estimates should be on average less than 0.4 percentage points for a 5-year period.[35] From fiscal years 2020 through 2025, BLS met its goal but reported annual benchmark revisions close to or higher than this level in some years.[36] For example, in fiscal year 2025, BLS released an annual benchmark revision that revised its March 2024 employment estimate downward 0.4 percentage points, its largest benchmark revision since 2009. In its preliminary annual benchmark revision release in September 2025, the agency also estimated it would revise its March 2025 employment estimate downward by 0.6 percentage points. In the final annual benchmark revision released in February 2026, BLS revised its March 2025 estimate downward by 0.5 percentage points.

BLS Faces Growing Risks to Data Quality as Fewer Individuals and Employers Respond to Surveys, Among Other Challenges

Stakeholders we interviewed generally agreed that the Jobs Report data are sufficiently accurate, useful, and timely for informing economic policy and business decisions. However, most stakeholders cited lower survey response rates as a challenge that has either already affected the accuracy and usefulness of the report or is expected to do so in the future. From 2015 to 2025, response rates to the household survey fell from 87 to 69 percent, and response rates to the establishment survey fell from 62 to 42 percent (see fig. 6).[37] Response rates to both surveys fell substantially following the onset of the COVID-19 pandemic in January 2020, and have not returned to their pre-COVID-19 pandemic levels.

Note: We analyzed response rates to the household and establishment surveys from October 2015 to September 2025. This was the latest full fiscal year of data available at the time of our analysis. BLS measures response rates to the establishment survey based on the percentage of responses it receives in time for the third release of the data. Percentages are rounded to the nearest whole number.

^aHousehold survey refers to the Current Population Survey.

^bEstablishment survey refers to the Current Employment Statistics program.

 

How Do Lower Survey Response Rates Generally Affect Data Accuracy? Statisticians measure survey response rates—the proportion of a survey sample that responds. Lower survey response rates can affect data accuracy in two principal ways: 1.    Precision of survey estimates. Lower response rates result in smaller sample sizes, which in turn reduce precision in survey estimates and make it harder to distinguish important patterns or trends from random fluctuations. 2.    Nonresponse bias. If those who respond to the survey (e.g., households or employers) differ systematically from those who do not respond, survey data will no longer represent the intended population. This phenomenon (called “nonresponse bias”) can introduce additional errors that can push survey estimates in a certain direction, causing the estimates to be higher or lower than the true value. Statisticians can use computational adjustments to account for differences between survey respondents and nonrespondents to reduce the effect of nonresponse bias. However, the extent to which such adjustments eliminate nonresponse bias depends on how well they capture these differences. Source: GAO analysis of Federal Committee on Statistical Methodology, A Framework for Data Quality, FCSM-20-04 (Sept. 2020) and Jennifer H. Madans, et al. Best Practices for Nonresponse Bias Reporting, FCSM-23-01 Nonresponse Bias Subcommittee, Federal Committee on Statistical Methodology, June 2023.  |  GAO‑26‑107538

Several stakeholders explained that the decline in responses to the household and establishment survey reflects broader ongoing issues affecting surveys more generally.[38] Lower response rates can affect data accuracy by leading to less precise estimates and introducing nonresponse bias (see sidebar).[39]

Household survey response rate challenges. As discussed previously, BLS met its goal for the precision of its national unemployment rate estimates from fiscal years 2020 through 2025. However, we found that declining survey response rates over this period nevertheless resulted in modest declines in overall precision, as measured by the statistical margin of error. BLS officials told us that these trends led them to revise the agency’s precision requirement for the household survey in April 2025 to permit a higher margin of error. This means the agency will be able to meet its performance goal with less precise estimates. Going forward, the agency will require that the survey be sufficiently precise for a 0.3 percentage point change in the unemployment rate to be statistically significant, an increase from the previous 0.2 percent threshold.

Reduced precision can erode the usefulness of data like the unemployment rate for making specific policy and business decisions, according to several of the stakeholders we interviewed. For example, two stakeholders said that when buying and selling financial assets, stock traders and other financial market participants can make decisions based on small changes to unemployment. These data users would like to have confidence that a change as small as 0.1 percentage points reflects a meaningful change in employment conditions. Should the precision of the household survey continue to decline, such changes are more likely to be viewed as due to chance and are less likely to reflect meaningful trends.

BLS has an annual usefulness goal for the household survey to release a certain number of monthly estimates from the survey to data users.[40] For example, for fiscal year 2020, BLS’s goal was to release on average at least 14,900 monthly estimates. These estimates include data on unemployment rates and other labor force information for various demographic subgroups.

While declining precision of the household survey poses a risk to usefulness, BLS met its usefulness goal consistently from fiscal years 2020 through 2025. It also raised its goal for the number of estimates to release over this period, from 14,900 to 19,900, reflecting expansions in the number of demographic groups included and additional data on telework.

Studies we reviewed found that declining response rates resulted in limited nonresponse bias in key estimates from the household survey such as the national unemployment rate.[41] However, these studies have not examined whether nonresponse bias has impacted estimates since 2021. In one study, Census staff found that the sharp decline in survey response at the start of the COVID-19 pandemic did not lead to a meaningful effect on the estimated unemployment rate.[42] A study authored by BLS staff found similar results.[43] The Census study also found that lower response rates from 2017 to 2019 did not significantly bias estimates of the unemployment rate. This result was supported by an academic study that we reviewed.[44]

BLS and Census are taking several steps to address declining response rates and provide data users with more information on the household survey’s accuracy. First, they are planning to introduce a new response method that would allow respondents to complete the survey questionnaire online rather than through an in-person or telephone interview.[45] Second, Census staff are working on analyzing the effect of nonresponse bias on unemployment and employment from 2020 to 2024. They expect to release the results of their analysis in May 2026. Third, BLS is also expanding its level of transparency regarding the precision of the household survey estimates. Specifically, BLS plans to release data for calculating margins of error for almost all data items that they release with the Jobs Report by the end of fiscal year 2026.[46]

Establishment survey response rate challenges. Lower response rates have posed challenges to BLS’s ability to meet the establishment survey’s annual usefulness goal. BLS’s goal is to release a certain number of monthly estimates from the survey to data users.[47] For example, for fiscal year 2020, BLS’s goal was to release at least 25,450 monthly estimates.[48] These estimates include data on employment, hours, and earnings for various industries. The agency met its usefulness goal in fiscal year 2020, fiscal year 2021, and fiscal year 2023, but it did not meet its goal in fiscal year 2022, fiscal year 2024, and fiscal year 2025.[49]

BLS officials told us they revisit their usefulness goal annually following a review of the estimates, with the aim of releasing information on as many industries as possible while meeting the agency’s quality and confidentiality standards.[50] Due to falling response rates and other factors that have resulted in smaller sample sizes, BLS reduced the number of estimates it releases to meet its standards and also lowered its related performance goal. Specifically, from fiscal year 2020 through fiscal year 2025 BLS lowered its goal for the number of estimates it releases from the establishment survey from 25,450 to 22,900. For instance, BLS no longer reports employment information for some detailed manufacturing industries, such as breakfast cereal manufacturing. BLS officials told us that declining manufacturing employment has also contributed to smaller sample sizes for such detailed industries over time.

To help address these challenges, BLS has taken steps to maintain the usefulness of the establishment survey. For example, BLS staff participated in a cross-country effort to update the North American Industry Classification System, which BLS and other government agencies use to classify industries.[51] This effort resulted in changes in 2023 that helped the establishment survey more accurately capture retail trade and e-commerce activity, according to BLS officials. In addition, in 2025, BLS reviewed the industry-level estimates it did not previously publish due to lower sample sizes in the establishment survey and published some of that data.

Stakeholders and research identified three additional challenges that can affect the quality of the Jobs Report data: (1) occasional large revisions to the employment data, (2) out-of-date population information, and (3) difficulties measuring alternative work arrangements.

BLS’s Model for Estimating Business Openings and Closures For the Employment Situation report (Jobs Report), BLS uses a statistical model to account for employment changes resulting from business openings and closures in the establishment survey data. It calls this model the “net birth-death model.” BLS uses this model because it would otherwise be difficult for its survey estimates to reflect employment changes associated with business openings (births) and closures (deaths). It takes time for BLS to include new businesses into its survey sample, or remove businesses after they have closed, due to lags in when BLS receives the administrative data it uses to capture this information.  Unaddressed, this limitation would have greater impact on the accuracy of the estimates from the establishment survey because openings and closures are important contributors to employment changes. BLS uses past data and statistical methods in the net birth-death model to estimate the effect of business openings and closures on monthly employment. Source: GAO analysis of Bureau of Labor Statistics (BLS) information.  |  GAO‑26‑107538

Occasional large revisions in the establishment survey. Most stakeholders we interviewed said that large revisions to the establishment survey can be a challenge. For example, one stakeholder said that uncertainty created by larger revisions can delay decision-making because data users must wait longer to more confidently understand trends in the data. Larger monthly revisions led this stakeholder to analyze the estimates using a 3-month rolling average to smooth out volatility and incorporate private data sources when monitoring employment trends.

For the annual benchmark revision, one source of revisions is forecast error from BLS’s model for estimating business openings and closures, known as “the net birth-death model” (see sidebar). The model improves the measurement of national payroll employment during stable economic conditions. However, the model can result in the agency making larger annual benchmark revisions during periods of sudden economic changes, such as the recession of 2007 to 2009, according to BLS documentation and some stakeholders. This is because the model has relied on historical data to forecast how business openings and closures contribute to monthly changes in employment.[52]

Despite the potential for error, 12 of the 14 stakeholders said the overall accuracy of the Jobs Report data, including the size of revisions to establishment survey data, was acceptable for informing economic policy and business decisions. The revision process is designed to provide users with initial estimates soon after the month is over, followed by revised estimates as more complete workforce data become available. Some stakeholders said the revision process itself is a strength because it provides more accurate data over time. BLS has also met its own goals related to the size of the revisions. BLS officials told us they continue to study improvements that could help reduce the size of revisions, and the agency adjusted its net birth-death model in February 2026 in an effort to make its employment estimates more accurate.[53]

Out-of-date population information. Out-of-date population information resulted in understated population and employment levels from the household survey, according to three studies that we reviewed and recent BLS analysis. BLS reported that the household survey understated the population of people living in the U.S. in 2024 by 2.9 million (or 1.1 percent), likely due to challenges that Census had capturing recent immigration in its projections.[54] As a result, the household survey underestimated the total number of employed individuals in the Jobs Report for various demographic groups in 2024; however, it had a limited effect on unemployment rates, according to the studies we reviewed. Census updated its methodology for the 2025 population projections to better reflect immigration by incorporating a wider range of data sources.

Difficulties measuring alternative work arrangements. Three research studies we reviewed and some stakeholders we interviewed said that the household survey questionnaire mismeasures alternative work arrangements, such as informal and gig work.[55] Individuals do not always report these arrangements, particularly if they only work a few hours a week. This undercounting can affect the measurement of employment rates, but studies we reviewed suggest that the impact of mismeasurement of informal work on overall employment may be limited. We previously found that disagreements exist among researchers about the magnitude of the errors in the household survey related to gig work and other nonstandard work arrangements.[56] BLS officials told us they are currently exploring methods to better capture data on alternative work arrangements, such as through a supplemental survey or potentially adding a new question to the household survey.

BLS Does Not Have a Plan to Address Gaps in Obtaining Input from Technical Experts and Data Users on Efforts to Improve Data Quality

While BLS obtains some input from data users and technical experts, gaps exist in obtaining regular feedback to inform improvements to the Jobs Report. BLS officials said they obtain information on users’ needs from various sources. For example, program staff receive inquiries via phone or e-mail regarding the data. BLS also relies on media accounts of labor market issues and public research to understand users’ needs. Program staff use this information to assess whether there are patterns of inquiry that warrant a change. In addition, BLS officials said they obtain feedback from technical experts when presenting at research conferences and publishing in peer-reviewed publications.

However, in February 2025, the Departments of Labor and Commerce eliminated three data user and expert advisory committees following an Executive Order after they determined that the committees had fulfilled their purpose.[57] BLS had relied on these committees to obtain regular feedback on key improvement efforts for the Jobs Report and other data products and to obtain the views of technical experts on BLS’s statistical practices (see table 1).

Advisory committee	Purpose and membership	Example of improvement effort discussed at meetings
BLS Data Users Advisory Committee	Provided advice to BLS from the perspective of data users Composed of knowledgeable data users from labor, business, government, and other sectors of the U.S. economy	Changes to a supplement to the household surveya that BLS uses to collect additional data on alternative work arrangements
BLS Technical Advisory Committee	Provided advice to BLS on best practices related to economics, statistics, and survey design Composed of technical experts	Plans to modernize the household survey and to add an online response method to address declining response rates
Department of Commerce Federal Economic Statistics Advisory Committee	Provided advice to BLS, Census, and the Bureau of Economic Analysis on statistical best practices related to federal economic data Composed of technical experts and knowledgeable data users	New household survey questions to measure telework and other impacts of the COVID-19 pandemic Changes to BLS’s net birth-death model that accounts for business openings and closures in the establishment surveyb

Source: GAO analysis of Bureau of Labor Statistics (BLS) and Department of Commerce information.  |  GAO‑26‑107538

Note: The Departments of Labor and Commerce eliminated the BLS Data Users Advisory Committee, the BLS Technical Advisory Committee, and the Federal Economic Statistics Advisory Committee in 2025 in response to Exec. Order No. 14,217, Commencing the Reduction of the Federal Bureaucracy, 90 Fed. Reg. 10,577 (Feb. 25, 2025).

^aHousehold survey refers to the Current Population Survey.

^bEstablishment survey refers to the Current Employment Statistics program.

While the advisory committees had overlapping scope and membership, each served a distinct advisory role for BLS prior to their elimination, according to BLS and Department of Commerce documentation. For example, the Federal Economic Statistics Advisory Committee provided BLS, Census, and the Bureau of Economic Analysis input related to various economic data. In comparison, the BLS Data Users Advisory Committee and BLS Technical Advisory Committee provided input to BLS alone related to the agency’s broader portfolio of data products, including noneconomic data such as occupational safety and health statistics.

The committees also had different membership bases, with the BLS Technical Advisory Committee and the Federal Economic Statistics Advisory Committee composed primarily of researchers with advanced backgrounds in statistics, economics, and survey design and the BLS Data Users Advisory Committee composed of a broader pool of users. These users included researchers as well as others, such as representatives of employment services and news organizations.

BLS officials acknowledged that gaps exist in their ability to obtain regular external feedback for the Jobs Report without these committees.[58] For example, BLS may not be able to readily obtain information on the needs of private sector and other users that BLS officials do not routinely interact with. They also said they will have difficulty obtaining certain high-level technical advice that was previously provided by the BLS Technical Advisory Committee and the Federal Economic Statistics Advisory Committee. BLS officials also indicated that the committees played an important role and provided a formal public-facing mechanism for feedback on their improvement efforts. BLS officials said the agency is not developing a plan to address the gaps due to limited resources and said they are prioritizing efforts to maintain data quality.

While BLS may be able to obtain additional external input using other sources of information, it is not clear whether those sources will meet its need for a broad range of input from users and technical experts. For example, one BLS statistician explained that they planned to attend research conferences to obtain input on changes to the net birth-death model that could reduce annual benchmark revisions. They explained that the conference participants could provide general feedback, but they did not expect that participants would have the methodological expertise that would have been represented on the BLS Technical Advisory Committee.

OMB policy requires that federal statistical agencies, including BLS, regularly seek input from a broad range of data users, and BLS policy requires that program methodologies are based on sound technical practices.[59] In addition, federal standards for internal control call for agencies to obtain relevant and quality information from external parties to help them achieve their objectives.[60] Proactively developing a plan, including time frames for implementation, to address gaps in external input could help BLS ensure that it obtains relevant and timely information to make informed decisions about the Jobs Report data and acts as needed to enhance data quality. Doing so would improve the likelihood that future changes to the data meet users’ needs, respond to risks, and benefit from needed technical input, which could help BLS maintain the quality of the data over time.

BLS and Census Assessed Lower Survey Response Rates but Shared Limited Information About What They Learned with Data Users

BLS and Census have assessed the effects of lower survey response rates on the accuracy of the data in the Jobs Report, but the agencies have shared limited information about what they have learned from these assessments with data users. We compared BLS’s and Census’s assessment efforts from 2020 to 2025 to relevant OMB and BLS requirements for assessing response rates and communicating the findings of the assessments to users. We assessed the agencies’ efforts for the household and establishment surveys separately. Figure 7 shows the extent to which the agencies met the requirements.

Note: We did not assess the extent to which BLS measured effects of methodological changes on the establishment survey because there were no response rate-related changes from 2020 to 2025, according to BLS officials. We identified federal requirements for our evaluation from the following OMB and BLS policies: OMB, Fundamental Responsibilities of Recognized Statistical Agencies and Units, 89 Fed. Reg. 82,453 (Oct. 11, 2024) (codified at 5 C.F.R. pt. 1321); Standards and Guidelines for Statistical Surveys, 71 Fed. Reg. 55,522 (Sept. 22, 2006); Update of Statistical Policy Directive No. 3: Compilation, Release, and Evaluation of Principal Federal Economic Indicators—Changing Timing of Public Comments by Employees of the Executive Branch, 89 Fed. Reg. 11,873 (Feb. 15, 2024); BLS, Commissioner’s Order No. 30-7 (Mar. 2024) and Program and Reviewer Checklist of BLS Principles for BLS Quality Reviews.

^aHousehold survey refers to the Current Population Survey.

^bEstablishment survey refers to the Current Employment Statistics program.

BLS and Census Monitored and Measured the Effects of Lower Survey Response Rates on Key Employment Data in the Jobs Report

From 2020 to 2025, BLS and Census fully met OMB and BLS requirements for assessing response rates for the household and establishment surveys for the Jobs Report. Further, both agencies are taking additional steps to understand the effects of survey response rates on data quality. OMB directs agencies that conduct statistical surveys to measure response rates and conduct nonresponse bias analyses when response rates suggest the potential for bias to occur.[61] In addition, BLS policy requires programs to conduct nonresponse bias analyses when response rates fall below 80 percent.[62] Response rates to both surveys fell over this period following the COVID-19 pandemic to levels that would require a nonresponse bias assessment under these requirements.[63] Consistent with OMB’s and BLS’s requirements, both agencies have monitored response rates and evaluated whether lower response rates since 2020 impacted key data in the report. For example:

·         Household survey studies. As we previously discussed, BLS and Census staff published two nonresponse bias analyses that found limited effects of the COVID-19 pandemic on the accuracy of the national unemployment rate.[64]

·         Establishment survey studies. BLS has annually analyzed nonresponse bias in total payroll employment. Specifically, BLS performs an error analysis, in which it measures how nonresponse bias and other different types of statistical error affect the annual benchmark revision.[65] BLS conducts this analysis by comparing the original survey-based employment estimates to those based on more comprehensive administrative employment records that the agency uses for its annual benchmark revision. To measure nonresponse bias, BLS calculates the difference in payroll employment between employers who responded to the survey and the full sample selected for the establishment survey, including those that did not respond.[66] BLS has used this analysis to understand broadly how nonresponse and other types of statistical error affect the annual benchmark revision, according to BLS officials.[67] However, BLS officials told us the analysis also has methodological limitations that have prevented the agency from releasing the results publicly. (We discuss these methodological limitations and how they have affected what the agency has released in more detail below.)

BLS and Census have also taken steps to obtain additional information on how nonresponse bias has affected the Jobs Report data from both surveys. For the household survey, Census officials told us that the agency is coordinating with BLS on a new analysis that examines whether declining response rates from 2020 to 2024 resulted in any bias. In addition, they said that Census anticipates releasing this new study in May 2026 and plans to perform a new nonresponse bias analysis every two or three years. For the establishment survey, BLS told us it has contracted with an outside entity to conduct a new nonresponse bias study for the establishment survey. The new study is intended to help identify improvements to the survey. BLS expects this study to be completed in 2026.

BLS also fully met relevant OMB requirements for assessing survey response rates that are specifically for agencies that produce principal federal economic indicators like the Jobs Report.[68] For example, BLS has submitted periodic evaluations of survey response rates and other issues to OMB for both the household and establishment surveys.[69] BLS also evaluated a change it made in 2023 to the household survey methodology to use more current response rate information.[70]

Agencies Have Shared Limited Information About the Effects of Lower Survey Response Rates on Data Accuracy but Have Plans to Provide Additional Information

BLS and Census have fully met OMB’s requirements and have adhered to BLS’s policies for providing data users with information about response rates and potential sources of error in the Jobs Report. However, the agencies have shared limited information with data users about what they have learned from their assessments of nonresponse bias specifically. As a result, users of the Jobs Report may not understand how, if at all, lower response rates to the surveys have affected data accuracy since 2020.

Information on response rates and potential sources of error. BLS has publicly reported response rates for the household survey, which data users can download from the agency’s website. The BLS website has also provided users with response rates for the establishment survey. In addition, BLS’s and Census’s documentation for the Jobs Report and the surveys have identified survey nonresponse as a potential source of error in the data.[71]

Conclusions of nonresponse bias assessments. BLS and Census have shared limited information with Jobs Report users about what the agencies have learned from their nonresponse bias assessments of the household survey. In addition, the information that has been available has been outdated. For example, until July 2025, BLS’s webpage with response rate information on the household survey and links to available nonresponse bias assessments did not include a 2024 Census study on nonresponse bias during the COVID-19 pandemic.[72] Further, Census did not release results from a study it conducted in 2022 comparing the characteristics of household survey respondents and nonrespondents. BLS has also not provided users with an overall summary of the extent to which lower survey response rates since 2020 have resulted in nonresponse bias in the unemployment rate or other Jobs Report data from the household survey.[73]

BLS and Census officials told us that they have not provided more information about nonresponse bias in the household survey because research is still evolving. They also said past research findings may be specific to certain time periods and therefore may not currently be representative. BLS officials said they plan to include a link to Census’s new nonresponse bias analysis on its webpage once the analysis is completed. They also told us they will review the results and determine whether there are concrete findings to summarize and share with data users.

We also found that BLS has provided data users with information on overall statistical error in the establishment survey data for the Jobs Report, as measured by the size of the annual benchmark revision. As part of its benchmarking process, BLS releases documentation that describes the methodology and reports the revision to employment numbers for the U.S. as a whole and by industry. The documentation also describes how forecast errors in BLS’s net birth-death model contributed to the revision and actions BLS has taken to address identified survey errors.[74]

However, BLS’s documentation for the annual benchmark revision and other key documentation for the establishment survey have provided data users with limited information regarding how the agency assesses nonresponse bias or what it has found. Specifically:

·         Key online documentation for the report and the establishment survey do not explain BLS’s methodology for the error analysis that it has used to understand how nonresponse bias and other types of statistical error have affected the annual benchmark revision to total payroll employment.

·         This documentation also lacks information on what BLS found from its error analyses from 2020 to 2024.

·         BLS referenced some of its conclusions from its error analyses in 2025 after reporting large annual benchmark revisions.[75] However, BLS did not explain how it formed its conclusions. It also did not provide users with a summary assessment of overall nonresponse bias on payroll employment in the establishment survey.

BLS officials told us that their error analyses have provided the agency with useful information about nonresponse bias and other types of statistical error. For example, BLS officials used the results to investigate revisions for specific industries using raw microdata for the establishment survey. However, the agency has not released the results publicly because of methodological limitations. Specifically, BLS experienced challenges matching the administrative records for the analyses with the original survey responses. The data also have not undergone BLS’s data quality review, such as automated screening and manual review of potential outliers. BLS officials said the results could lead users to draw inaccurate conclusions.

BLS officials explained that it has not provided data users with more information on nonresponse bias in the establishment survey because it would take considerable resources to address these methodological limitations and release the results of the error analyses to the public. However, officials said the agency intends to provide data users with more information after its new nonresponse study is completed in 2026. In addition, the new study is intended to address the methodological limitations of BLS’s error analyses. The study is also intended to result in a reproduceable framework for summarizing the effect of nonresponse bias that BLS could then release to data users on a recurring basis. However, it is not yet clear what information BLS will share publicly. BLS has also not shared publicly the methodology used in its error analyses for assessing nonresponse bias and other sources of statistical error.

OMB policies require BLS to publicly release both documentation with information necessary for evaluating survey results and its evaluations of data quality, and to have a dissemination plan that provides for equivalent, timely access to information to all data users.[76] BLS policy also requires that the agency’s technical documentation include a summary assessment of all sources of error, including direction (i.e., positive or negative) and order of magnitude of errors, with a special focus on key estimates.[77]

By updating its public documentation to explain its current methodology for assessing nonresponse bias in the establishment survey, BLS would be more transparent about how the agency monitors a potentially important source of error in the Jobs Report data that can also affect the annual benchmark revision. In addition, publishing at least summary-level information on how survey nonresponse has affected employment and other estimates from the survey would further strengthen that transparency and help users of the Jobs Report understand how lower response rates have affected the data. This, in turn, could increase users’ confidence in the data and minimize the possibility of users making decisions based on incomplete information. This summary could be drawn from the agency’s new nonresponse bias study under contract or through other means. Greater transparency could also help maintain users’ trust in the revision process and the data’s accuracy.

Agency Efforts and Options Identified by Stakeholders to Address Survey Challenges and Enhance Data Quality Involve Trade-Offs

Agency Efforts That Could Enhance Data Quality

BLS and Census are engaged in efforts aimed at addressing survey challenges and enhancing the quality of the Jobs Report data (see table 2).[78]

Agency effort	How the effort is intended to address survey challengesa	Key considerations and trade-offs, according to agency officials or stakeholders
Household surveya modernization: Multiyear effort to modernize the household survey, which will include adding an online response method scheduled to be introduced in 2027.	·          Increase response rates ·          Manage data collection costs	·          Could increase response rates and manage data collection costs. ·          Could allow BLS and Census to field a larger survey sample. ·          Requires reviews and testing to understand potential effects of online data collection. ·          Funding constraints and other challenges have limited field testing and may delay full implementation.
Improvements to the model estimating business openings and closures: BLS introduced technical improvements in 2026 to the statistical model it uses to account for employment changes resulting from business openings and closures, which would otherwise be difficult to account for in survey estimates.	·          Reduce annual benchmark revisions to the employment estimates from the establishment surveya	·          Expected to reduce error in the model, which should result in more accurate monthly employment estimates and smaller annual benchmark revisions.

Source: GAO analysis of interviews with Bureau of Labor Statistics (BLS) officials, Census officials, and selected stakeholders.  |  GAO‑26‑107538

^aThe Jobs Report includes data from two surveys: the Current Population Survey (the household survey) and the Current Employment Statistics program (the establishment survey).

Household Survey Modernization

BLS and Census are engaged in a multiyear effort to modernize the household survey, which will include adding an online response method. BLS and Census began developing the online response method in fiscal year 2023. It is scheduled to be introduced in 2027.[79]

The key goals of the household survey modernization are to increase response rates and manage data collection costs, including through implementation of the new online response method, according to BLS and Census. As part of the broader modernization effort, BLS and Census are pursuing other updates to data collection processes, including improved cost-tracking, enhanced training for field staff, and data-driven approaches for conducting survey outreach.[80]

Seven stakeholders who discussed the online response method generally anticipated that it would help increase response rates or reduce data collection costs. In particular, three of the stakeholders said the online response method would help to increase survey participation. Six of the stakeholders said it would lower costs, for instance by reducing the need for Census interviewers to collect information from respondents.

Stakeholders and agency officials also anticipated that the planned online response method may have additional benefits. For instance, by potentially lowering data collection costs, the online response method could allow the household survey to field a larger sample, according to two stakeholders. Census officials also said the online response method could improve the quality of the household survey data by reaching households less likely to respond to in-person or telephone surveys. Finally, one stakeholder said the online response method could provide an opportunity for BLS and Census to adapt survey questions and field additional surveys quickly to respond to emerging data needs, such as those that arose during the COVID-19 pandemic.[81]

Stakeholders also identified potential challenges associated with the online response method. Online surveys can lead to differences in responses compared to in-person or telephone surveys, which could affect survey estimates and limit the comparability of data over time, according to five stakeholders. Moreover, two stakeholders raised concerns that the population of individuals who respond to the online survey may have different characteristics compared to the general population. For instance, one stakeholder said that online survey respondents would be more likely to be younger, in the labor force, and more familiar with technology, among other differences.

BLS and Census are conducting extensive reviews and testing to understand the effects of the online response method, which is a key reason for the multiyear implementation timeline. However, agency officials said that recent funding constraints, staffing reductions, and delays related to the 2025 government shutdown have slowed development of the online response method and other parts of the modernization effort, limited the scope of their field tests, and may delay the full implementation of the online response method.

Improvements to the Model Estimating Business Openings and Closures

As discussed previously, in February 2026, BLS made technical improvements to its net birth-death model that estimates business openings and closures each month for the establishment survey. These technical improvements could help reduce annual benchmark revisions to the payroll employment estimates by incorporating information from the establishment survey into the model to increase the model’s sensitivity to current economic conditions.[82]

BLS officials said they anticipate that these technical improvements will be effective at increasing the accuracy of the estimates of business openings and closures that are generated by the model. Specifically, officials said research has shown that the improvements will reduce error in the model, which should result in more accurate monthly estimates in the Jobs Report and smaller annual benchmark revisions.[83]

Potential Options That Could Enhance Data Quality

Stakeholders, researchers, and BLS officials identified potential options aimed at addressing survey challenges and enhancing data quality. Each option involves trade-offs, such as potential improvements to data quality, increased financial costs, and additional respondent burden (see table 3).

Option	How the option is intended to address survey challengesa	Key considerations and trade-offs, according to agency officials, stakeholders, or research
Blended data: Incorporating alternative data sources into the household survey and establishment surveya estimates, such as unemployment insurance data or private payroll data.	·          Mitigate effects of lower response rates ·          Reduce monthly revisions and annual benchmark revisions to establishment survey estimates ·          Improve data collection on alternative work arrangements, such as “gig work”	·          Could help increase the accuracy and granularity of the Jobs Report data. ·          BLS and Census officials cited challenges around effectiveness, feasibility, and costs. ·          Would require significant research and testing.
Mandatory surveys: Making the household and establishment surveys mandatory for the individuals and employers randomly selected to participate.	·          Increase response rates	·          Research has shown that mandatory surveys have higher response rates. ·          BLS officials raised concerns about potential negative reactions to enforcing mandatory survey participation. ·          BLS officials said it would require an act of Congress to make the surveys mandatory.
Monetary incentives for household survey: Offering monetary incentives for household survey respondents.	·          Increase response rates	·          Research has shown that offering monetary incentives can increase survey participation. ·          Would involve costs to pay incentives, but costs may be offset by needing fewer interviewers to follow up with nonrespondents.
More resources for administering household survey: Providing more resources, such as additional staff or funding, to BLS and Census to help with administering the household survey.	·          Increase response rates and sample size	·          Could allow for additional follow-up with nonrespondents to increase survey response rates. ·          Could allow for a larger initial survey sample to account for drop-off in participation over time. ·          Would involve increased budgetary costs for BLS and Census to administer the survey.
Statistical weighting adjustments for establishment survey estimates: Developing methods for adjusting statistical survey weights to account for differences in the characteristics of responding employers compared to the broader sample of selected employers.	·          Reduce monthly revisions and annual benchmark revisions to establishment survey estimates	·          Could improve accuracy of employment estimates by accounting for fluctuations in the characteristics of employers who respond to the survey over time.
Updating household survey questions: Updating the household survey questions to better capture information on alternative work arrangements (e.g., informal work and gig work).	·          Improve data collection on alternative work arrangements	·          Updating the household survey’s questions could help address limitations in gathering data on alternative work arrangements, for instance by including a specific question about “side jobs.” ·          BLS officials said such updates would require significant testing, financial costs, and cutting other survey questions to limit respondent burden. ·          BLS is exploring other methods to capture this information.

Source: GAO analysis of interviews with Bureau of Labor Statistics (BLS) officials, Census officials, selected stakeholders, and review of selected research. | GAO‑26‑107538

^aThe Jobs Report includes data from two surveys: the Current Population Survey (the household survey) and the Current Employment Statistics program (the establishment survey).

Blended Data

A potential approach for addressing a range of survey challenges could be to use blended data. Using blended data would involve incorporating alternative data sources into the household and establishment survey estimates. Blended data could involve the use of government administrative data from federal or state sources (such as tax records or unemployment insurance data), private data from payroll processing companies, or other federal and nongovernmental economic data. This would be intended to address challenges related to the effects of lower survey response rates on data quality, large monthly revisions and annual benchmark revisions to the establishment survey data, and difficulty gathering information on alternative work arrangements (e.g., informal work or gig work).

 

 

 

 

The Jobs and Employment Data Exchange The U.S. Chamber of Commerce Foundation is leading a public-private initiative, the Jobs and Employment Data Exchange, focused on standardizing, consolidating, and automating employer data reporting to government agencies. BLS and two states were partners in the initiative’s pilot test, which demonstrated how states could collect unemployment insurance and related data from employers in a more streamlined process, and then report the data to BLS, according to U.S. Chamber of Commerce Foundation staff. U.S. Chamber of Commerce Foundation staff said this initiative could lead to fundamental changes in how employment data is gathered and reported to government agencies. They said this could provide agencies with more timely data on a range of employment measures from a larger number of employers, while reducing costs to employers and the government. BLS officials did not expect that this initiative would replace the Current Employment Statistics program (the establishment survey) but said it could improve the usefulness and granularity of other employment data gathered by BLS and other federal statistical agencies. Source: GAO analysis of interviews with Bureau of Labor Statistics (BLS) officials and U.S. Chamber of Commerce Foundation staff and associated documentation.  |  GAO‑26‑107538

Stakeholders and research have highlighted different ways blended data could be used. For instance, two stakeholders said that updating state unemployment insurance systems and reporting processes could allow BLS to obtain employment data from a larger number of employers, almost in real time, which could help improve the accuracy and granularity of the Jobs Report estimates.[84] This could build on steps that BLS and others are currently taking to standardize and streamline reporting of unemployment insurance data (see sidebar).

Furthermore, data from private payroll processing companies could supplement the data gathered in the establishment survey, according to two stakeholders. In addition, one research study suggested that BLS could improve the accuracy of its initial estimates of payroll employment by using various federal and nongovernmental data sources on employment and other economic factors, including data from the household survey.[85] Finally, another study suggested that various data sources, including tax data and nongovernmental data, could be used to improve estimates of alternative work arrangements.[86]

However, BLS and Census officials cited several challenges associated with blended data and generally did not believe it would be effective or feasible to directly incorporate alternative data sources into the Jobs Report estimates. For instance, BLS and Census officials said that government administrative data are generally not timely enough to use for calculating the report’s key measures, such as total payroll employment growth or the unemployment rate. In addition, BLS officials said that administrative data, such as state unemployment insurance records, can contain errors and have differing scopes and definitions compared to BLS’s survey data. According to officials, this would create obstacles to using it in the monthly estimates.

Census officials said they are researching other ways to use administrative data to help improve response rates for the household survey and account for potential nonresponse bias in the survey’s estimates. Specifically, this research involves using administrative data to identify who is most likely to respond to the survey and through which methods (e.g., in person, via telephone), as well as identify the characteristics of survey nonrespondents. This could allow Census to develop more effective strategies to contact potential survey respondents and to adjust survey estimates to account for nonresponse bias, respectively.[87]

Stakeholders and agency officials cited several additional considerations related to the use of blended data. For instance, this approach would require extensive research and testing. It could also involve significant financial costs, such as costs to update state unemployment insurance systems or to acquire data from private payroll processing companies. In addition, private data sources are not always designed in a way that matches the needs of federal statistical agencies. Finally, if the private sector source of the data were to go out of business, the data would no longer be available.

Mandatory Surveys

A potential approach for addressing lower response rates in the household and establishment surveys would be to designate these surveys as mandatory for the individuals and employers that Census and BLS randomly select to participate in them.[88] While both of these surveys are voluntary at the federal level, some states require employers to participate in the establishment survey.[89]

Mandatory surveys have been shown to have higher response rates compared to voluntary surveys, according to Census officials and one stakeholder.[90] The stakeholder said that making the household and establishment surveys mandatory could be an effective approach for increasing response rates for both surveys. However, BLS officials said that states where the establishment survey is mandatory do not have noticeably higher response rates than states where it is not mandatory. Officials said this may be due to a lack of enforcement of those state requirements.

BLS officials also said that BLS surveys have been voluntary in nature since the agency’s founding. Furthermore, officials did not believe that an increase in response rates from enforcing mandatory participation in the household or establishment survey would be worth the potential for negative reactions to such a requirement. In addition, they said making the surveys mandatory would require an act of Congress.

Monetary Incentives for Household Survey

Another potential approach for addressing declining response rates to the household survey is to offer monetary incentives to survey respondents, according to research cited by one stakeholder and Census officials.[91] BLS and Census are currently studying the use of monetary incentives as part of the online response method they are developing for the household survey, according to agency officials.[92] However, BLS and Census officials said they were currently uncertain regarding the potential effectiveness and feasibility of this approach for increasing survey response rates, as they had only recently begun field testing.

This approach would involve financial costs to the agency to pay survey respondents for participating, according to BLS officials. However, BLS and Census officials said it is possible that these costs would be balanced by reduced overall data collection costs if more respondents respond to the initial survey request, which could reduce the need for costly survey follow-up by Census interviewers. Finally, offering monetary incentives for the household survey would require approval from the Office of Management and Budget, according to BLS officials.

More Resources for Administering Household Survey

Another potential approach for addressing declining response rates to the household survey would be to provide more resources, such as additional funding or staff, to BLS and Census for administering the survey.[93]

Providing more resources for administering the household survey would be an effective way to increase response rates and boost the survey’s sample size, according to one stakeholder. In particular, the stakeholder said that additional funding and staff capacity would allow Census interviewers to perform additional follow-up with nonrespondents, especially through in-person interviews, which could help increase response rates. The stakeholder also said these additional resources would allow BLS and Census to increase the number of individuals sampled in the first month of the survey to account for a drop-off in participation over subsequent months.[94]

This approach would involve increased budgetary costs for BLS and Census to administer the household survey.

Statistical Weighting Adjustments for Establishment Survey Estimates

BLS is researching methods to adjust its establishment survey estimates to account for fluctuations in the characteristics of employers who respond to the survey over time, which could help reduce monthly revisions and annual benchmark revisions.

Survey Weighting Statisticians use survey weighting to develop estimates that more accurately reflect the population a survey seeks to represent. This process involves assigning values (or “weights”) to each response to account for important differences in the characteristics of the groups that are sampled or that respond to a survey compared to the broader population those groups are meant to represent. For example, if employers in certain industries are underrepresented in a survey’s sample or among those who respond to the survey, those employers’ responses would be assigned larger weights to ensure the survey’s estimates reflect the broader population. Survey weighting can improve the accuracy of estimates produced from survey data but is subject to limitations. For example, survey weights can only account for differences that are observed in the data, and remaining differences may bias survey estimates. In addition, weighting can result in less precise survey estimates even when it reduces bias. Source: GAO analysis of Jelke Bethlehem, “Weighting,” in Encyclopedia of Survey Research Methods, (Vol. 0, pp. 958-960). Sage Publications, Inc., Jennifer H. Madans, et al., Best Practices for Nonresponse Bias Reporting, FCSM-23-01 Nonresponse Bias Subcommittee, Federal Committee on Statistical Methodology, June 2023, and Bureau of Labor Statistics documentation.  |  GAO‑26‑107538

BLS officials said that taking such steps could help improve the accuracy of monthly employment estimates. Specifically, BLS researchers are developing methods for adjusting statistical survey weights to account for differences in the characteristics of responding employers compared to the broader sample of selected employers. For example, BLS officials said that there can be differences in employer response rates by industry over time.

Officials said the most significant challenge to implementation is identifying correct adjustments to account for changes in the characteristics of responding employers without introducing new error into the employment estimates.

Updating Household Survey Questions

Updating the household survey’s questions could help address limitations in gathering information on alternative work arrangements (e.g., informal work and gig work), according to one stakeholder and two studies in our literature review.[95] For example, one study suggested that the household survey could clarify that respondents should include any type of informal work activity in their answers about employment, or the survey could include a specific question about “side jobs.”[96] BLS officials were uncertain about the potential effectiveness and feasibility of this approach, explaining that alternative work arrangements can be difficult to measure. In addition, BLS does not have an official definition for what is considered “gig work.”[97]

BLS officials also said that updating the household survey questions would require extensive testing, involve significant financial costs, and require BLS to cut other portions of the survey to limit respondent burden. However, BLS is exploring methods to better capture these data through an occasional supplement to the household survey.[98] Furthermore, BLS is field testing the use of a potential follow-up question in its planned online response method, which would ask those respondents who reported no employment if they worked for even as little as one hour during the specified week.

Conclusions

The monthly Jobs Report provides critical information on the health of the nation’s economy and is a key resource for guiding economic policy and business decisions. However, key challenges, including lower response rates for the household and establishment surveys that are the basis for the report present growing risks to BLS’s and Census’s ability to maintain the data’s accuracy and usefulness.

BLS and Census have taken steps to mitigate these and other challenges. For example, the agencies are working to add an online response method for the household survey aimed at improving response rates. BLS has also made adjustments to its model for estimating business openings and closures for the establishment survey, which could help reduce annual benchmark revisions. The agencies are also exploring additional options to adapt, including offering monetary incentives to increase household survey participation and using statistical weighting techniques to improve the accuracy of and reduce revisions to the establishment survey estimates.

However, BLS does not have a plan to address gaps in its ability to obtain regular input from knowledgeable external stakeholders after the Departments of Commerce and Labor eliminated three advisory committees that BLS previously relied on to ensure that it took needed actions to maintain or improve data quality. In addition, BLS has not yet explained to the public its methodology for assessing nonresponse bias and other sources of statistical error on the establishment survey. It also has not published a summary assessment of nonresponse bias on this survey, despite concerns about falling response rates and occasional large revisions.

Without taking steps to address these issues, BLS risks that the Jobs Report data will not adequately meet the needs of data users and that data users may not fully understand how to interpret key data in the report. Taking additional actions in these areas would help BLS ensure that the Jobs Report remains a timely and trusted resource for measuring the health of the economy and helping policymakers and businesses make informed decisions.

Recommendations for Executive Action

We are making the following three recommendations to the Department of Labor:

The Secretary of Labor should ensure that the Commissioner of BLS, in consultation with the Director of the Census Bureau as appropriate, develops and implements a plan to address gaps in BLS’s ability to obtain regular external input on changes to the Employment Situation report (Jobs Report) data. The plan should include timelines for implementation. (Recommendation 1)

The Secretary of Labor should ensure that the Commissioner of BLS updates the public documentation for the Current Employment Statistics program (establishment survey) to explain BLS’s current methodology for assessing nonresponse bias and other sources of statistical error on key survey estimates, including any limitations in its methodology. (Recommendation 2)

The Secretary of Labor should ensure that the Commissioner of BLS publishes a summary assessment of nonresponse bias on the Current Employment Statistics program (establishment survey). The assessment should include the magnitude and direction of any bias on key survey estimates, such as total payroll employment growth. (Recommendation 3)

Agency Comments and Our Evaluation

We provided a draft of our report for review and comment to BLS, the Department of Commerce’s Census Bureau and Bureau of Economic Analysis, the Department of the Treasury, and the Board of Governors of the Federal Reserve System. We received written comments from BLS, which are reproduced in appendix III and summarized below. BLS and Census provided technical comments, which we incorporated, as appropriate. The Bureau of Economic Analysis and the Board of Governors of the Federal Reserve System did not have any comments on the report. Treasury did not provide comments on the report.

In its written comments, BLS agreed with the two recommendations that were related to the establishment survey. BLS stated it planned to implement these recommendations by drawing on the outcomes of its new nonresponse bias study under contract. We discussed this study in our second research objective.

BLS did not disagree with our recommendation to address gaps in its ability to obtain regular external input on changes to the Jobs Report data. BLS noted that it has existing channels to obtain stakeholder input but did not specify whether it has used these channels to specifically discuss the Jobs Report. It also said it would work within applicable guidance to identify additional opportunities to formalize and expand stakeholder input, including through advisory mechanisms where appropriate. We will continue to monitor BLS’s implementation of this recommendation.

We are sending copies of this report to the appropriate congressional committees, the Secretary of Labor, the Secretary of Commerce, the Secretary of the Treasury, the Chair of the Board of Governors of the Federal Reserve System, and other interested parties. In addition, the report is available at no charge on the GAO website at https://www.gao.gov.

If you or your staff have any questions about this report, please contact us at costat@gao.gov or hoffmanme@gao.gov. Contact points for our Offices of Congressional Relations and Media Relations may be found on the last page of this report. GAO staff who made significant contributions to this report are listed in appendix IV.

Thomas Costa
Director
Education, Workforce, and Income Security

Michael Hoffman
Chief Economist
Applied Research and Methods

The objectives of this review were to examine (1) the extent to which the employment data in the Employment Situation report (Jobs Report) meet users’ needs for accurate, useful, and timely information and challenges the Bureau of Labor Statistics (BLS) faces in producing these data; (2) the extent to which BLS and the Census Bureau have followed federal requirements for assessing survey response rates and for communicating the findings of these assessments to data users; and (3) what potential options BLS, Census, stakeholders, and research have identified to address survey challenges, and what considerations are associated with each option.

Overall Summary of Methods

To address our first objective, we assessed the accuracy, usefulness, and timeliness of the Jobs Report data based on relevant dimensions from the Federal Committee on Statistical Methodology data quality framework:[99]

·         We assessed BLS’s annual performance goals for the two surveys that collect the data used in the Jobs Report—the Current Population Survey (the “household survey”) and the Current Employment Statistics program (the “establishment survey”). We determined the extent to which BLS met these goals from fiscal years 2020 through 2025.

·         We reviewed literature published from January 2020 to September 2025 that examined the accuracy of the employment data in either the household survey or the establishment survey.

·         We interviewed 14 stakeholders with expertise on the data and methods in the Jobs Report to obtain their views on the accuracy, usefulness, and timeliness of the employment data for informing economic policy and business decisions.[100] We also discussed current and emerging challenges faced by data users, BLS, and Census related to data quality (which we refer to as “survey challenges”).

·         We assessed the extent to which BLS’s processes for obtaining user and technical expert input on the Jobs Report data met BLS and Office of Management and Budget (OMB) policies and federal standards for internal control.[101]

To address our second objective, we reviewed relevant information on survey response rates and assessed BLS’s and Census’s processes for assessing the effects of response rates on the Jobs Report data’s accuracy. We also evaluated the extent to which these processes met relevant OMB and BLS policies and relevant federal standards for internal control.[102]

To address our third objective, we interviewed BLS and Census officials, as well as our selected stakeholders, to identify agency initiatives and potential options to address challenges with data quality. We also identified potential options suggested by studies from our literature review. Furthermore, we reviewed relevant agency research and documentation, as well as research shared by stakeholders, to further understand these options and key considerations. We also analyzed BLS budget data for the household and establishment surveys from fiscal years 2020 through 2025. We do not endorse any particular option.

Review of BLS Performance Goals and Budget Data

We assessed whether BLS met its performance goals for the accuracy, usefulness, and timeliness of the household and establishment surveys from fiscal years 2020 through 2025 and the extent to which the agency adjusted these goals over this period.[103] We determined whether BLS met its performance goals by comparing data we obtained from Congressional Budget Justifications, Agency Management Plans, and Operating Plans to the agency’s reported goals.[104] For time periods in which this information was not available through these sources, we obtained data directly from BLS documentation and officials and calculated performance results, as needed. We also analyzed trends in survey response rates, which BLS uses to monitor data collection.[105] We obtained response rates from the BLS website.

To provide additional context for performance results, we analyzed data that BLS uses to calculate its accuracy goals. In particular, we analyzed trends in margins of error to the national unemployment rate and revisions to payroll employment. We obtained these data from BLS officials and documentation.[106]

We also analyzed BLS budget obligations data for the household and establishment surveys from fiscal years 2020 through 2025 to corroborate information we heard from agency officials and stakeholders. We obtained these data from BLS officials. The budget obligations for these surveys reflect final amounts incurred during the fiscal year and do not include centralized charges for items such as rent, security charges, and other information that are recorded at the budget activity level, according to BLS officials.

We assessed the reliability of these data by reviewing relevant documentation, interviewing knowledgeable officials, and conducting electronic data testing.[107] We found the data to be sufficiently reliable for determining the extent to which BLS met its performance goals for the Jobs Report and reviewing recent survey response rates and agency budget trends.

Literature Review

To understand how survey challenges may have affected the accuracy of the Jobs Report data and to identify potential options to address any accuracy issues, we conducted a literature review. Specifically, we reviewed government and academic studies that examined the accuracy of the employment data in either the household survey or the establishment survey that were published from January 2020 to September 2025.

We identified studies from several sources. In April 2025 we conducted an extensive search of pertinent scholarly databases, such as EconLit, Social SciSearch, and Scopus. We also conducted web searches for working papers published by BLS, Census, and Federal Reserve researchers and those affiliated with the National Bureau of Economic Research. In August and September 2025, we updated our searches to identify studies published in 2025 within our scope. Through this process, we identified 280 potentially relevant studies.

We screened all 280 studies using a multistage approach. During the initial stage, we reviewed the studies’ abstracts to determine whether they examined the accuracy of the employment data in either the household survey or the establishment survey. Multiple GAO staff, including economists and statisticians, reviewed each study abstract for relevance. If we could not determine relevance from the abstract alone, we completed a review of the full text. We removed studies that did not evaluate the accuracy of the household or establishment surveys from further review. From an initial pool of 280 studies, we selected 31 studies that met our criteria.

We completed structured reviews of these 31 studies’ methodologies and findings.[108] We developed a data collection instrument that described each of the selected studies. Multiple GAO staff, including economists and statisticians, reviewed each study in its entirety and completed the data collection instrument at this stage. Each instrument required the reviewers to summarize information from the study and determine whether the study’s methodology was appropriate and sufficiently rigorous for the purposes of our review. The data collection instrument included the study’s objectives and findings, a description of the methodology, any limitations for how the results could be interpreted, and what options, if any, were discussed to address identified accuracy issues.

Interviews with Stakeholders

We interviewed 14 stakeholders with expertise on the data and methods in the Jobs Report. We obtained their views on the accuracy, usefulness, and timeliness of the employment data for informing economic policy and business decisions and to identify options for addressing survey challenges.[109] The stakeholders included representatives of organizations that use these data and other knowledgeable individuals. Specifically, we selected a nongeneralizable sample of stakeholders from the following categories: (1) public policy decision-makers and advisors, (2) banks and investment firms, (3) business associations, (4) former BLS commissioners, and (5) researchers with expertise related to the household or establishment surveys. We also considered the following factors in our selection process: (1) membership or organizational representation on one of BLS’s advisory committees at the time of selection, (2) recommendations from relevant subject matter experts, and (3) selections to obtain a range of institutional perspectives.[110] Table 4 lists the stakeholders we interviewed.

Stakeholder/Category

Public policy decision-makers and advisors

Board of Governors of the Federal Reserve System

Congressional Budget Office

Department of the Treasury

Federal Reserve Bank of Chicago

Federal Reserve Bank of Cleveland

Banks and investment firms

Goldman Sachs

JPMorgan Chase

Vanguard

Business associations

National Federation of Independent Business

U.S. Chamber of Commerce

Two former Bureau of Labor Statistics commissioners

Two researchers with expertise related to the household or establishment surveysa

Source: GAO.  |  GAO‑26‑107538

^aHousehold survey refers to the Current Population Survey. Establishment survey refers to the Current Employment Statistics program.

The interviews included two parts:

1.    We asked stakeholders to provide closed-ended responses to 10 statements regarding the accuracy, usefulness, and timeliness of the data in the Jobs Report, using a five-point scale ranging from “strongly agree” to “strongly disagree” (see table 5 for a summary of the stakeholder responses).[111]

2.    We used a semi-structured interview protocol to gather information on the specific strengths and limitations of the Jobs Report data, including current and emerging challenges faced by data users, BLS, and Census related to data quality.[112] Specifically, this portion of the interview covered: (1) stakeholders’ backgrounds and experience with the Jobs Report data, (2) how their institutions or economic policymakers and businesses generally use the data, (3) strengths and limitations of the household and establishment survey data, (4) emerging data challenges that could affect the quality of the Jobs Report, and (5) BLS efforts and additional options to address survey challenges, including the anticipated effectiveness and key considerations associated with the efforts and options (e.g., effects on data quality, feasibility, costs, and other trade-offs).[113]

In January 2026, we followed up with each of the stakeholders to see if they had additional observations to share. The nine stakeholders who responded indicated that their views on the quality of the data had not substantively changed since the time of our interviews.

Statement	Strongly Agree	Agree	Neither Agree Nor Disagree	Disagree	Strongly Disagree	Do Not Know/No Basis to Judge
The information and employment data provided in the report meet [my institution’s needs/my agency’s needs/economic policymakers’ and business needs].	8	6	0	0	0	0
The report’s analysis and methods are based on accepted statistical practices.	13	1	0	0	0	0
The report’s overall accuracy and reliability (e.g., sampling and nonsampling errors, size of revisions) are acceptable for [my institution’s needs/my agency’s needs/economic policymakers’ and business needs].	5	7	2	0	0	0
The information and employment data in the report are provided at a level of granularity that is adequate for [my institution’s needs/my agency’s needs/economic policymakers’ and business needs].	5	5	3	0	0	1
The documentation for the Employment Situation data meets [my institution’s needs/my agency’s needs/economic policymakers’ and business needs].	11	3	0	0	0	0
The Employment Situation data and documentation can be obtained in understandable forms and formats.	9	4	0	1	0	0
The information in the report and technical documentation is adequate for understanding sources of error, key limitations, assessing appropriate uses of the data, and interpreting the results.	8	5	0	1	0	0
The information in the report is sufficiently timely to meet [my institution’s needs/my agency’s needs/economic policymakers’ and business needs].	8	6	0	0	0	0
It is clear when the report (including scheduled revisions) will be released, and releases are issued on schedule.	13	1	0	0	0	0
The report is the best available source for current data on national employment and unemployment.	12	2	0	0	0	0

Source: GAO analysis of information provided by stakeholders.  |  GAO‑26‑107538

Note: We used minor variations in the wording of some statements based on stakeholder type. Specifically, these statements asked whether the data met “my agency’s needs” for federal agency stakeholders, “my institution’s needs” for other stakeholder institutions, and “economic policymakers’ and business needs” for the stakeholder categories of former BLS commissioners and researchers with expertise related to the household or establishment surveys.

Evaluation of Agency Assessments and Communication on Survey Response Rates

We evaluated the extent to which BLS and Census have followed federal requirements for assessing survey response rates and for communicating the findings of these assessments to data users. To identify the requirements that we used in our evaluation, we reviewed relevant BLS and OMB policies and interviewed BLS and Census officials.[114] We synthesized the results into six unique requirements, removing requirements that were either not applicable or out of scope: (1) monitoring response rates and nonresponse bias, (2) submitting performance evaluations to OMB, (3) measuring effects of methodological changes, (4) providing response rates to data users, (5) acknowledging potential sources of error, and (6) summarizing conclusions of assessments.

To assess the extent to which BLS and Census have followed these six requirements, we reviewed the data and documentation on survey response rates that BLS and Census publish, as well as internal and public BLS and Census assessments of the effects of response rates on data quality. Our review included key documentation that BLS produces for data users of the Jobs Report, including the report’s technical documentation and the BLS Handbook of Methods, as well as other information available on the BLS and Census websites. We also interviewed BLS and Census officials to understand their processes for reviewing, assessing, and communicating information on response rates to data users.

We determined whether BLS and Census had fully met, partially met, or not at all met each of the six requirements.[115] We determined the requirement was fully met if available evidence demonstrated all aspects of the requirement; partially met if available evidence demonstrated some, but not all, aspects of the requirement; and not at all met if available evidence did not demonstrate any aspects of the requirement. We assessed the agencies’ efforts for the household and establishment surveys separately. We also considered whether these processes met relevant federal standards for internal control.

To understand how survey challenges may have affected the accuracy of the Jobs Report data and to identify potential options to address any accuracy issues, we reviewed the following 31 studies.

Abraham, Katharine G., John Haltiwanger, Claire Hou, Kristin Sandusky, and James R. Spletzer. “Reconciling Survey and Administrative Measures of Self-Employment.” Journal of Labor Economics, vol. 39, no. 4 (2021).

Abraham, Katharine G., Brad Hershbein, Susan N. Houseman, and Beth C. Truesdale. “The Independent Contractor Workforce: New Evidence on its Size and Composition and Ways to Improve its Measurement in Household Surveys.” ILR Review, vol. 77, no. 3 (2024).

Ahn, Joo Hie and James D. Hamilton. “Measuring Labor-Force Participation and the Incidence and Duration of Unemployment.” Review of Economic Dynamics, vol. 44 (2022).

Barnichon, Regis and Winnie Yee. “Adjusting the Unemployment Thermometer.” Federal Reserve Bank of San Francisco Economic Letter, No. 2020-27 (2020).

Bernhardt, Robert, David Munro, and Erin. L. Wolcott. “How Does the Dramatic Rise in Nonresponse in the Current Population Survey Impact Labor Market Indicators?” Journal of Applied Econometrics, vol. 39 (2024).

Bonnéry, Daniel, Yang Cheng, and Partha Lahiri. “An Evaluation of Design-Based Properties of Different Composite Estimators.” Working Paper, April 2020. https://arxiv.org/pdf/1811.12249.

Bracha, Anat and Mary A. Burke. “Informal Work and Official Employment Statistics: What’s Missing?” Research Department Working Papers No. 23-15, Federal Reserve Bank of Boston, September 2023.

Cai, Yixia and Dean Baker. “Masking Real Unemployment: The Overall and Racial Impact of Survey Non-Response on Measured Labor Market Outcomes.” Working Paper No. 150, Institute for New Economic Thinking, March 2021.

Coglianese, John, Seth Murray, and Christopher J. Nekarda. “Harmonized Population and Labor Force Statistics.” Finance and Economics Discussion Series 2025-057, Board of Governors of the Federal Reserve System, July 2025.

Dorfman, Jeffrey, Wenying Li, and Jingfang Zhang. “Forecasting Revisions to the U.S. Jobs Report Data.” The B.E. Journal of Macroeconomics (2025).

Edelberg, Wendy, Olivia Howard, Eileen Powell, and Tara Watson. Higher New Census Population Estimates Will Affect the Employment Report. Brookings Institution, February 2025.

Edelberg, Wendy and Tara Watson. New Immigration Estimates Help Make Sense of the Pace of Employment. The Hamilton Project, Brookings Institution, March 2024.

Eggleston, Jonathan, Yarissa Gonzalez, Carl Lieberman, Tim Trudell, and John Voorheis. “Incorporating Administrative Data in Survey Weights for the Basic Monthly Current Population Survey.” Discussion Paper CES 24-02, U.S. Census Bureau, January 2024.

Feng, Shuaizhang, Yingyao Hu, and Jiandong Sun. “Rotation Group Bias and the Persistence of Misclassification Errors in the Current Population Surveys.” Econometric Reviews, vol. 41, no. 9 (2022).

Grieves, Chris, Steve Mance, and Collin Witt. Predicting the Effects of Business Births and Deaths on the Current Employment Statistics Survey: Using Sample Information to Minimize Coverage Error. Bureau of Labor Statistics, December 2023.

Heffetz, Ori and Daniel Reeves. “Measuring Unemployment in Crisis: Effects of COVID-19 on Potential Biases in the CPS.” NBER Working Paper No. 28310, December 2020.

Hudson, Nicole, Jeannine Mercurio, and Jurgen Kropf. “The Challenges of Seasonal Adjustment for the Current Employment Statistics Survey During the COVID-19 Pandemic.” Monthly Labor Review, Bureau of Labor Statistics, May 2022.

Lahiri, P. and Bogong T. Li. “Linear Empirical Bayes Prediction of Employment Growth Rates Using the U.S. Current Employment Statistics Survey.” In Strategic Management, Decision Theory, and Decisions Science, ed. B.K. Sinha and S.B. Bagchi. Springer, Singapore, 2021.

Leduc, Sylvain, Luiz E. Oliveira, and Caroline M. Paulson. “Do Low Survey Response Rates Threaten Data Dependence?” Federal Reserve Bank of San Francisco Economic Letter, No. 2025-07 (2025).

McIllece, Justin J. COVID-19 and the Current Population Survey: Response Rates and Estimation Bias. Bureau of Labor Statistics, October 2020.

McIllece, Justin J. Optimizing the Current Population Survey Composite Estimator. Bureau of Labor Statistics, October 2022.

McIllece, Justin J. Propensity-Adjusted Raking with Applications to Current Population Survey Nonresponse. Bureau of Labor Statistics, 2023.

Rhein, Bradley D. Imputing for Extraordinary Sample Events: A Story of Targeted Donor Pools and Administrative Data. Bureau of Labor Statistics, December 2023.

Robertson, Kenneth W. A Better Benchmark Process for the U.S. Current Employment Statistics Survey. Bureau of Labor Statistics, December 2021.

Robertson, Kenneth W. BLS Business Surveys in the Wake of COVID-19 – Changes to Data Collection, Imputation, and Estimation. Bureau of Labor Statistics, December 2021.

Severen, Christopher. “Delayed Sampling of Recent Immigrants in the Current Population Survey.” Federal Reserve Bank of Philadelphia Research Briefs (2025).

Shibata, Ippei. “Reassessing Classification Errors in the Analysis of Labor Market Dynamics.” Labour Economics, vol. 78 (2022).

U.S. Census Bureau. Current Population Survey Nonresponse Bias Analysis as of April 2022. Washington, D.C.: October 2022.

Ward, Jason M. and Kathryn Anne Edwards. “CPS Nonresponse During the COVID-19 Pandemic: Explanations, Extent, and Effects.” Labour Economics, vol. 72 (2021).

Willis, Jonathan L. and Tao Zha. “What Accounts for the Growing Divergence Between Employment Measures?” Federal Reserve Bank of Atlanta Policy Hub, No. 6-2024 (2024).

Witt, Collin. Testing of Weight Smoothing Models in the Current Employment Statistics Survey with SAS and R. Bureau of Labor Statistics, August 2023.

GAO Contacts

Thomas Costa, costat@gao.gov

Michael Hoffman, hoffmanme@gao.gov

Staff Acknowledgments

In addition to the contacts named above, Nisha R. Hazra (Assistant Director), Carl Nadler (Analyst in Charge), Natalie Duncombe, and Abinash Mohanty made key contributions to this report. Also contributing to this report were Dana Hopings, Kelsey Kreider, Carly McCann, Michael Naretta, David Reed, Jason Rodriguez Masi, Lindsay Shapray, Jared Smith, Meg Sommerfeld, Almeta Spencer, Will Stupski, Rebecca Kuhlmann Taylor, and Walter Vance.

GAO’s Mission

The Government Accountability Office, the audit, evaluation, and investigative arm of Congress, exists to support Congress in meeting its constitutional responsibilities and to help improve the performance and accountability of the federal government for the American people. GAO examines the use of public funds; evaluates federal programs and policies; and provides analyses, recommendations, and other assistance to help Congress make informed oversight, policy, and funding decisions. GAO’s commitment to good government is reflected in its core values of accountability, integrity, and reliability.

Obtaining Copies of GAO Reports and Testimony

The fastest and easiest way to obtain copies of GAO documents at no cost is through our website. Each weekday afternoon, GAO posts on its website newly released reports, testimony, and correspondence. You can also subscribe to GAO’s email updates to receive notification of newly posted products.

Order by Phone

The price of each GAO publication reflects GAO’s actual cost of production and distribution and depends on the number of pages in the publication and whether the publication is printed in color or black and white. Pricing and ordering information is posted on GAO’s website, https://www.gao.gov/ordering.htm.

Place orders by calling (202) 512-6000, toll free (866) 801-7077, or
TDD (202) 512-2537.

Orders may be paid for using American Express, Discover Card, MasterCard, Visa, check, or money order. Call for additional information.

Connect with GAO

Connect with GAO on X, LinkedIn, Instagram, and YouTube.
Subscribe to our Email Updates. Listen to our Podcasts.
Visit GAO on the web at https://www.gao.gov.

To Report Fraud, Waste, and Abuse in Federal Programs

Contact FraudNet:

Website: https://www.gao.gov/about/what-gao-does/fraudnet

Automated answering system: (800) 424-5454

Media Relations

Sarah Kaczmarek, Managing Director, Media@gao.gov

Congressional Relations

David A. Powner, Acting Managing Director, CongRel@gao.gov

General Inquiries

https://www.gao.gov/about/contact-us